Commit Graph

398 Commits

Author SHA1 Message Date
adrianeboyd faaa832518 Generalize handling of tokenizer special cases (#4259)
* Generalize handling of tokenizer special cases

Handle tokenizer special cases more generally by using the Matcher
internally to match special cases after the affix/token_match
tokenization is complete.

Instead of only matching special cases while processing balanced or
nearly balanced prefixes and suffixes, this recognizes special cases in
a wider range of contexts:

* Allows arbitrary numbers of prefixes/affixes around special cases
* Allows special cases separated by infixes

Existing tests/settings that couldn't be preserved as before:

* The emoticon '")' is no longer a supported special case
* The emoticon ':)' in "example:)" is a false positive again

When merged with #4258 (or the relevant cache bugfix), the affix and
token_match properties should be modified to flush and reload all
special cases to use the updated internal tokenization with the Matcher.

* Remove accidentally added test case

* Really remove accidentally added test

* Reload special cases when necessary

Reload special cases when affixes or token_match are modified. Skip
reloading during initialization.

* Update error code number

* Fix offset and whitespace in Matcher special cases

* Fix offset bugs when merging and splitting tokens
* Set final whitespace on final token in inserted special case

* Improve cache flushing in tokenizer

* Separate cache and specials memory (temporarily)
* Flush cache when adding special cases
* Repeated `self._cache = PreshMap()` and `self._specials = PreshMap()`
are necessary due to this bug:
https://github.com/explosion/preshed/issues/21

* Remove reinitialized PreshMaps on cache flush

* Update UD bin scripts

* Update imports for `bin/`
* Add all currently supported languages
* Update subtok merger for new Matcher validation
* Modify blinded check to look at tokens instead of lemmas (for corpora
with tokens but not lemmas like Telugu)

* Use special Matcher only for cases with affixes

* Reinsert specials cache checks during normal tokenization for special
cases as much as possible
  * Additionally include specials cache checks while splitting on infixes
  * Since the special Matcher needs consistent affix-only tokenization
    for the special cases themselves, introduce the argument
    `with_special_cases` in order to do tokenization with or without
    specials cache checks
* After normal tokenization, postprocess with special cases Matcher for
special cases containing affixes

* Replace PhraseMatcher with Aho-Corasick

Replace PhraseMatcher with the Aho-Corasick algorithm over numpy arrays
of the hash values for the relevant attribute. The implementation is
based on FlashText.

The speed should be similar to the previous PhraseMatcher. It is now
possible to easily remove match IDs and matches don't go missing with
large keyword lists / vocabularies.

Fixes #4308.

* Restore support for pickling

* Fix internal keyword add/remove for numpy arrays

* Add test for #4248, clean up test

* Improve efficiency of special cases handling

* Use PhraseMatcher instead of Matcher
* Improve efficiency of merging/splitting special cases in document
  * Process merge/splits in one pass without repeated token shifting
  * Merge in place if no splits

* Update error message number

* Remove UD script modifications

Only used for timing/testing, should be a separate PR

* Remove final traces of UD script modifications

* Update UD bin scripts

* Update imports for `bin/`
* Add all currently supported languages
* Update subtok merger for new Matcher validation
* Modify blinded check to look at tokens instead of lemmas (for corpora
with tokens but not lemmas like Telugu)

* Add missing loop for match ID set in search loop

* Remove cruft in matching loop for partial matches

There was a bit of unnecessary code left over from FlashText in the
matching loop to handle partial token matches, which we don't have with
PhraseMatcher.

* Replace dict trie with MapStruct trie

* Fix how match ID hash is stored/added

* Update fix for match ID vocab

* Switch from map_get_unless_missing to map_get

* Switch from numpy array to Token.get_struct_attr

Access token attributes directly in Doc instead of making a copy of the
relevant values in a numpy array.

Add unsatisfactory warning for hash collision with reserved terminal
hash key. (Ideally it would change the reserved terminal hash and redo
the whole trie, but for now, I'm hoping there won't be collisions.)

* Restructure imports to export find_matches

* Implement full remove()

Remove unnecessary trie paths and free unused maps.

Parallel to Matcher, raise KeyError when attempting to remove a match ID
that has not been added.

* Switch to PhraseMatcher.find_matches

* Switch to local cdef functions for span filtering

* Switch special case reload threshold to variable

Refer to variable instead of hard-coded threshold

* Move more of special case retokenize to cdef nogil

Move as much of the special case retokenization to nogil as possible.

* Rewrap sort as stdsort for OS X

* Rewrap stdsort with specific types

* Switch to qsort

* Fix merge

* Improve cmp functions

* Fix realloc

* Fix realloc again

* Initialize span struct while retokenizing

* Temporarily skip retokenizing

* Revert "Move more of special case retokenize to cdef nogil"

This reverts commit 0b7e52c797.

* Revert "Switch to qsort"

This reverts commit a98d71a942.

* Fix specials check while caching

* Modify URL test with emoticons

The multiple suffix tests result in the emoticon `:>`, which is now
retokenized into one token as a special case after the suffixes are
split off.

* Refactor _apply_special_cases()

* Use cdef ints for span info used in multiple spots

* Modify _filter_special_spans() to prefer earlier

Parallel to #4414, modify _filter_special_spans() so that the earlier
span is preferred for overlapping spans of the same length.

* Replace MatchStruct with Entity

Replace MatchStruct with Entity since the existing Entity struct is
nearly identical.

* Replace Entity with more general SpanC

* Replace MatchStruct with SpanC

* Add error in debug-data if no dev docs are available (see #4575)

* Update azure-pipelines.yml

* Revert "Update azure-pipelines.yml"

This reverts commit ed1060cf59.

* Use latest wasabi

* Reorganise install_requires

* add dframcy to universe.json (#4580)

* Update universe.json [ci skip]

* Fix multiprocessing for as_tuples=True (#4582)

* Fix conllu script (#4579)

* force extensions to avoid clash between example scripts

* fix arg order and default file encoding

* add example config for conllu script

* newline

* move extension definitions to main function

* few more encodings fixes

* Add load_from_docbin example [ci skip]

TODO: upload the file somewhere

* Update README.md

* Add warnings about 3.8 (resolves #4593) [ci skip]

* Fixed typo: Added space between "recognize" and "various" (#4600)

* Fix DocBin.merge() example (#4599)

* Replace function registries with catalogue (#4584)

* Replace functions registries with catalogue

* Update __init__.py

* Fix test

* Revert unrelated flag [ci skip]

* Bugfix/dep matcher issue 4590 (#4601)

* add contributor agreement for prilopes

* add test for issue #4590

* fix on_match params for DependencyMacther (#4590)

* Minor updates to language example sentences (#4608)

* Add punctuation to Spanish example sentences

* Combine multilanguage examples for lang xx

* Add punctuation to nb examples

* Always realloc to a larger size

Avoid potential (unlikely) edge case and cymem error seen in #4604.

* Add error in debug-data if no dev docs are available (see #4575)

* Update debug-data for GoldCorpus / Example

* Ignore None label in misaligned NER data
2019-11-13 21:24:35 +01:00
Sofie Van Landeghem e48a09df4e Example class for training data (#4543)
* OrigAnnot class instead of gold.orig_annot list of zipped tuples

* from_orig to replace from_annot_tuples

* rename to RawAnnot

* some unit tests for GoldParse creation and internal format

* removing orig_annot and switching to lists instead of tuple

* rewriting tuples to use RawAnnot (+ debug statements, WIP)

* fix pop() changing the data

* small fixes

* pop-append fixes

* return RawAnnot for existing GoldParse to have uniform interface

* clean up imports

* fix merge_sents

* add unit test for 4402 with new structure (not working yet)

* introduce DocAnnot

* typo fixes

* add unit test for merge_sents

* rename from_orig to from_raw

* fixing unit tests

* fix nn parser

* read_annots to produce text, doc_annot pairs

* _make_golds fix

* rename golds_to_gold_annots

* small fixes

* fix encoding

* have golds_to_gold_annots use DocAnnot

* missed a spot

* merge_sents as function in DocAnnot

* allow specifying only part of the token-level annotations

* refactor with Example class + underlying dicts

* pipeline components to work with Example objects (wip)

* input checking

* fix yielding

* fix calls to update

* small fixes

* fix scorer unit test with new format

* fix kwargs order

* fixes for ud and conllu scripts

* fix reading data for conllu script

* add in proper errors (not fixed numbering yet to avoid merge conflicts)

* fixing few more small bugs

* fix EL script
2019-11-11 17:35:27 +01:00
Matthew Honnibal f8d740bfb1
Fix --gold-preproc train cli command (#4392)
* Fix get labels for textcat

* Fix char_embed for gpu

* Revert "Fix char_embed for gpu"

This reverts commit 055b9a9e85.

* Fix passing of cats in gold.pyx

* Revert "Match pop with append for training format (#4516)"

This reverts commit 8e7414dace.

* Fix popping gold parses

* Fix handling of cats in gold tuples

* Fix name

* Fix ner_multitask_objective script

* Add test for 4402
2019-10-27 21:58:50 +01:00
Sofie Van Landeghem 8e7414dace Match pop with append for training format (#4516)
* trying to fix script - not succesful yet

* match pop() with extend() to avoid changing the data

* few more pop-extend fixes

* reinsert deleted print statement

* fix print statement

* add last tested version

* append instead of extend

* add in few comments

* quick fix for 4402 + unit test

* fixing number of docs (not counting cats)

* more fixes

* fix len

* print tmp file instead of using data from examples dir

* print tmp file instead of using data from examples dir (2)
2019-10-27 16:01:32 +01:00
Ines Montani 5e59c9b3ee Fix unicode strings in examples [ci skip] 2019-10-18 18:47:59 +02:00
Sofie Van Landeghem 2d249a9502 KB extensions and better parsing of WikiData (#4375)
* fix overflow error on windows

* more documentation & logging fixes

* md fix

* 3 different limit parameters to play with execution time

* bug fixes directory locations

* small fixes

* exclude dev test articles from prior probabilities stats

* small fixes

* filtering wikidata entities, removing numeric and meta items

* adding aliases from wikidata also to the KB

* fix adding WD aliases

* adding also new aliases to previously added entities

* fixing comma's

* small doc fixes

* adding subclassof filtering

* append alias functionality in KB

* prevent appending the same entity-alias pair

* fix for appending WD aliases

* remove date filter

* remove unnecessary import

* small corrections and reformatting

* remove WD aliases for now (too slow)

* removing numeric entities from training and evaluation

* small fixes

* shortcut during prediction if there is only one candidate

* add counts and fscore logging, remove FP NER from evaluation

* fix entity_linker.predict to take docs instead of single sentences

* remove enumeration sentences from the WP dataset

* entity_linker.update to process full doc instead of single sentence

* spelling corrections and dump locations in readme

* NLP IO fix

* reading KB is unnecessary at the end of the pipeline

* small logging fix

* remove empty files
2019-10-14 12:28:53 +02:00
adrianeboyd d2d2baaf76 Revert training example edit from #4327 (#4403)
I think the original annotation was correct and this change also
unfortunately introduced a cycle into the dependency tree.
2019-10-10 17:00:26 +02:00
adrianeboyd 6f54e59fe7 Fix util.filter_spans() to prefer first span in overlapping sam… (#4414)
* Update util.filter_spans() to prefer earlier spans

* Add filter_spans test for first same-length span

* Update entity relation example to refer to util.filter_spans()
2019-10-10 17:00:03 +02:00
Ines Montani 52b5912dbf Tidy up [ci skip] 2019-10-02 12:05:59 +02:00
adrianeboyd d82241218a Make the default NER labels less model-specific [ci skip] (#4361) 2019-10-02 12:05:17 +02:00
Ines Montani b6670bf0c2 Use consistent spelling 2019-10-02 10:37:39 +02:00
Ines Montani 867e93aae2 Add Streamlit example [ci skip] 2019-10-02 01:21:20 +02:00
EarlGreyT 1e9e2d8aa1 fix typo in first token (#4327)
* fix typo in first token

The head of 'in' is review which has an offset of 4 and not 44

* added contributor agreement
2019-09-27 14:49:36 +02:00
adrianeboyd b5d999e510 Add textcat to train CLI (#4226)
* Add doc.cats to spacy.gold at the paragraph level

Support `doc.cats` as `"cats": [{"label": string, "value": number}]` in
the spacy JSON training format at the paragraph level.

* `spacy.gold.docs_to_json()` writes `docs.cats`

* `GoldCorpus` reads in cats in each `GoldParse`

* Update instances of gold_tuples to handle cats

Update iteration over gold_tuples / gold_parses to handle addition of
cats at the paragraph level.

* Add textcat to train CLI

* Add textcat options to train CLI
* Add textcat labels in `TextCategorizer.begin_training()`
* Add textcat evaluation to `Scorer`:
  * For binary exclusive classes with provided label: F1 for label
  * For 2+ exclusive classes: F1 macro average
  * For multilabel (not exclusive): ROC AUC macro average (currently
relying on sklearn)
* Provide user info on textcat evaluation settings, potential
incompatibilities
* Provide pipeline to Scorer in `Language.evaluate` for textcat config
* Customize train CLI output to include only metrics relevant to current
pipeline
* Add textcat evaluation to evaluate CLI

* Fix handling of unset arguments and config params

Fix handling of unset arguments and model confiug parameters in Scorer
initialization.

* Temporarily add sklearn requirement

* Remove sklearn version number

* Improve Scorer handling of models without textcats

* Fixing Scorer handling of models without textcats

* Update Scorer output for python 2.7

* Modify inf in Scorer for python 2.7

* Auto-format

Also make small adjustments to make auto-formatting with black easier and produce nicer results

* Move error message to Errors

* Update documentation

* Add cats to annotation JSON format [ci skip]

* Fix tpl flag and docs [ci skip]

* Switch to internal roc_auc_score

Switch to internal `roc_auc_score()` adapted from scikit-learn.

* Add AUCROCScore tests and improve errors/warnings

* Add tests for AUCROCScore and roc_auc_score
* Add missing error for only positive/negative values
* Remove unnecessary warnings and errors

* Make reduced roc_auc_score functions private

Because most of the checks and warnings have been stripped for the
internal functions and access is only intended through `ROCAUCScore`,
make the functions for roc_auc_score adapted from scikit-learn private.

* Check that data corresponds with multilabel flag

Check that the training instances correspond with the multilabel flag,
adding the multilabel flag if required.

* Add textcat score to early stopping check

* Add more checks to debug-data for textcat

* Add example training data for textcat

* Add more checks to textcat train CLI

* Check configuration when extending base model
* Fix typos

* Update textcat example data

* Provide licensing details and licenses for data
* Remove two labels with no positive instances from jigsaw-toxic-comment
data.


Co-authored-by: Ines Montani <ines@ines.io>
2019-09-15 22:31:31 +02:00
Sofie Van Landeghem 0b4b4f1819 Documentation for Entity Linking (#4065)
* document token ent_kb_id

* document span kb_id

* update pipeline documentation

* prior and context weights as bool's instead

* entitylinker api documentation

* drop for both models

* finish entitylinker documentation

* small fixes

* documentation for KB

* candidate documentation

* links to api pages in code

* small fix

* frequency examples as counts for consistency

* consistent documentation about tensors returned by predict

* add entity linking to usage 101

* add entity linking infobox and KB section to 101

* entity-linking in linguistic features

* small typo corrections

* training example and docs for entity_linker

* predefined nlp and kb

* revert back to similarity encodings for simplicity (for now)

* set prior probabilities to 0 when excluded

* code clean up

* bugfix: deleting kb ID from tokens when entities were removed

* refactor train el example to use either model or vocab

* pretrain_kb example for example kb generation

* add to training docs for KB + EL example scripts

* small fixes

* error numbering

* ensure the language of vocab and nlp stay consistent across serialization

* equality with =

* avoid conflict in errors file

* add error 151

* final adjustements to the train scripts - consistency

* update of goldparse documentation

* small corrections

* push commit

* typo fix

* add candidate API to kb documentation

* update API sidebar with EntityLinker and KnowledgeBase

* remove EL from 101 docs

* remove entity linker from 101 pipelines / rephrase

* custom el model instead of existing model

* set version to 2.2 for EL functionality

* update documentation for 2 CLI scripts
2019-09-12 11:38:34 +02:00
Sofie Van Landeghem 482c7cd1b9 pulling tqdm imports in functions to avoid bug (tmp fix) (#4263) 2019-09-09 16:32:11 +02:00
Ines Montani dad5621166 Tidy up and auto-format [ci skip] 2019-08-31 13:39:31 +02:00
adrianeboyd 82159b5c19 Updates/bugfixes for NER/IOB converters (#4186)
* Updates/bugfixes for NER/IOB converters

* Converter formats `ner` and `iob` use autodetect to choose a converter if
  possible

* `iob2json` is reverted to handle sentence-per-line data like
  `word1|pos1|ent1 word2|pos2|ent2`

  * Fix bug in `merge_sentences()` so the second sentence in each batch isn't
    skipped

* `conll_ner2json` is made more general so it can handle more formats with
  whitespace-separated columns

  * Supports all formats where the first column is the token and the final
    column is the IOB tag; if present, the second column is the POS tag

  * As in CoNLL 2003 NER, blank lines separate sentences, `-DOCSTART- -X- O O`
    separates documents

  * Add option for segmenting sentences (new flag `-s`)

  * Parser-based sentence segmentation with a provided model, otherwise with
    sentencizer (new option `-b` to specify model)

  * Can group sentences into documents with `n_sents` as long as sentence
    segmentation is available

  * Only applies automatic segmentation when there are no existing delimiters
    in the data

* Provide info about settings applied during conversion with warnings and
  suggestions if settings conflict or might not be not optimal.

* Add tests for common formats

* Add '(default)' back to docs for -c auto

* Add document count back to output

* Revert changes to converter output message

* Use explicit tabs in convert CLI test data

* Adjust/add messages for n_sents=1 default

* Add sample NER data to training examples

* Update README

* Add links in docs to example NER data

* Define msg within converters
2019-08-29 12:04:01 +02:00
adrianeboyd 8fe7bdd0fa Improve token pattern checking without validation (#4105)
* Fix typo in rule-based matching docs

* Improve token pattern checking without validation

Add more detailed token pattern checks without full JSON pattern validation and
provide more detailed error messages.

Addresses #4070 (also related: #4063, #4100).

* Check whether top-level attributes in patterns and attr for PhraseMatcher are
  in token pattern schema

* Check whether attribute value types are supported in general (as opposed to
  per attribute with full validation)

* Report various internal error types (OverflowError, AttributeError, KeyError)
  as ValueError with standard error messages

* Check for tagger/parser in PhraseMatcher pipeline for attributes TAG, POS,
  LEMMA, and DEP

* Add error messages with relevant details on how to use validate=True or nlp()
  instead of nlp.make_doc()

* Support attr=TEXT for PhraseMatcher

* Add NORM to schema

* Expand tests for pattern validation, Matcher, PhraseMatcher, and EntityRuler

* Remove unnecessary .keys()

* Rephrase error messages

* Add another type check to Matcher

Add another type check to Matcher for more understandable error messages
in some rare cases.

* Support phrase_matcher_attr=TEXT for EntityRuler

* Don't use spacy.errors in examples and bin scripts

* Fix error code

* Auto-format

Also try get Azure pipelines to finally start a build :(

* Update errors.py


Co-authored-by: Ines Montani <ines@ines.io>
Co-authored-by: Matthew Honnibal <honnibal+gh@gmail.com>
2019-08-21 14:00:37 +02:00
Sofie Van Landeghem 0ba1b5eebc CLI scripts for entity linking (wikipedia & generic) (#4091)
* document token ent_kb_id

* document span kb_id

* update pipeline documentation

* prior and context weights as bool's instead

* entitylinker api documentation

* drop for both models

* finish entitylinker documentation

* small fixes

* documentation for KB

* candidate documentation

* links to api pages in code

* small fix

* frequency examples as counts for consistency

* consistent documentation about tensors returned by predict

* add entity linking to usage 101

* add entity linking infobox and KB section to 101

* entity-linking in linguistic features

* small typo corrections

* training example and docs for entity_linker

* predefined nlp and kb

* revert back to similarity encodings for simplicity (for now)

* set prior probabilities to 0 when excluded

* code clean up

* bugfix: deleting kb ID from tokens when entities were removed

* refactor train el example to use either model or vocab

* pretrain_kb example for example kb generation

* add to training docs for KB + EL example scripts

* small fixes

* error numbering

* ensure the language of vocab and nlp stay consistent across serialization

* equality with =

* avoid conflict in errors file

* add error 151

* final adjustements to the train scripts - consistency

* update of goldparse documentation

* small corrections

* push commit

* turn kb_creator into CLI script (wip)

* proper parameters for training entity vectors

* wikidata pipeline split up into two executable scripts

* remove context_width

* move wikidata scripts in bin directory, remove old dummy script

* refine KB script with logs and preprocessing options

* small edits

* small improvements to logging of EL CLI script
2019-08-13 15:38:59 +02:00
svlandeg cd6c263fe4 format offsets 2019-07-23 11:31:29 +02:00
svlandeg 9f8c1e71a2 fix for Issue #4000 2019-07-22 13:34:12 +02:00
svlandeg dae8a21282 rename entity frequency 2019-07-19 17:40:28 +02:00
svlandeg 21176517a7 have gold.links correspond exactly to doc.ents 2019-07-19 12:36:15 +02:00
svlandeg e1213eaf6a use original gold object in get_loss function 2019-07-18 13:35:10 +02:00
svlandeg ec55d2fccd filter training data beforehand (+black formatting) 2019-07-18 10:22:24 +02:00
Ines Montani f2ea3e3ea2
Merge branch 'master' into feature/nel-wiki 2019-07-09 21:57:47 +02:00
Patrick Hogan 8c0586fd9c Update example and sign contributor agreement (#3916)
* Sign contributor agreement for askhogan

* Remove unneeded `seen_tokens` which is never used within the scope
2019-07-08 10:27:20 +02:00
svlandeg b7a0c9bf60 fixing the context/prior weight settings 2019-07-03 17:48:09 +02:00
svlandeg 8840d4b1b3 fix for context encoder optimizer 2019-07-03 13:35:36 +02:00
svlandeg 3420cbe496 small fixes 2019-07-03 10:25:51 +02:00
svlandeg 2d2dea9924 experiment with adding NER types to the feature vector 2019-06-29 14:52:36 +02:00
svlandeg c664f58246 adding prior probability as feature in the model 2019-06-28 16:22:58 +02:00
svlandeg 1c80b85241 fix tests 2019-06-28 08:59:23 +02:00
svlandeg 68a0662019 context encoder with Tok2Vec + linking model instead of cosine 2019-06-28 08:29:31 +02:00
svlandeg dbc53b9870 rename to KBEntryC 2019-06-26 15:55:26 +02:00
svlandeg 1de61f68d6 improve speed of prediction loop 2019-06-26 13:53:10 +02:00
svlandeg bee23cd8af try Tok2Vec instead of SpacyVectors 2019-06-25 16:09:22 +02:00
svlandeg b58bace84b small fixes 2019-06-24 10:55:04 +02:00
svlandeg a31648d28b further code cleanup 2019-06-19 09:15:43 +02:00
svlandeg 478305cd3f small tweaks and documentation 2019-06-18 18:38:09 +02:00
svlandeg 0d177c1146 clean up code, remove old code, move to bin 2019-06-18 13:20:40 +02:00
svlandeg ffae7d3555 sentence encoder only (removing article/mention encoder) 2019-06-18 00:05:47 +02:00
svlandeg 6332af40de baseline performances: oracle KB, random and prior prob 2019-06-17 14:39:40 +02:00
svlandeg 24db1392b9 reprocessing all of wikipedia for training data 2019-06-16 21:14:45 +02:00
svlandeg 81731907ba performance per entity type 2019-06-14 19:55:46 +02:00
svlandeg b312f2d0e7 redo training data to be independent of KB and entity-level instead of doc-level 2019-06-14 15:55:26 +02:00
svlandeg 0b04d142de regenerating KB 2019-06-13 22:32:56 +02:00
svlandeg 78dd3e11da write entity linking pipe to file and keep vocab consistent between kb and nlp 2019-06-13 16:25:39 +02:00
svlandeg b12001f368 small fixes 2019-06-12 22:05:53 +02:00